Check out our most recent work "On Covariate Shift of Latent Confounders in Imitation and Reinforcement Learning", where we talk about an important challenge of using expert data with hidden covariates. arxiv.org/pdf/2110.06539…

In case you missed our three papers in #Neurips2022, here they are again: 1. "RL with a Terminator" at the main conference: arxiv.org/abs/2205.15376 Guy Tennenholtz Nadav Merlis Lior Shani Uri Shalit shiemannor Gal Chechik

2. "Implementing RL Datacenter Congestion Control in NVIDIA NICs" at RL4RealLife and @ML4systems (spotlight) workshops: arxiv.org/abs/2207.02295 Benjamin Fuhrer yuval shpigelman Chen Tessler shiemannor Gal Chechik

3. "SoftTreeMax: Policy Gradient with Tree Search" at the @DeepRL workshop: arxiv.org/abs/2209.13966 shiemannor Gal Chechik

What’s Broken with RL Research and a Potential Fix: open.substack.com/pub/avivtamar/…

1/ Excited to share that our latest work, Conditional Adversarial Latent Models [CALM], has been accepted to ACM SIGGRAPH 2023. 🧵👇 #reinforcementlearning #animation #games #isaacgym #siggraph2023 NVIDIA AI

3/ What new capabilities does this unlock? Linearly interpolating between two motions (in the latent space) produces semantically meaningful transitions.

4/ We can leverage the semantically meaningful latent space for high-level style-conditioned policies. For instance, a high-level policy tasked with moving in a specified direction can be urged to use a specified style via a latent-space similarity reward (here, crouching).

5/ Combining the ability to control the style and direction a motion is produced -- we construct a finite state-machine to control the character, both the motion and direction it is performed.

6/ This enables solving unseen tasks in various forms, without training -- overcoming the need for meticulous reward/termination design.

7/ This work was made possible thanks to my amazing co-authors: Yoni Kasten, Yunrong Guo, Shie Mannor, Gal Chechik, and Jason Peng. More videos, a link to the paper, and code (coming real soon!): research.nvidia.com/labs/par/calm/

We released a multi-agent RL framework for network congestion control with the first public realistic network simulator! github.com/NVlabs/RLCC. Based on the amazing work of Benjamin Fuhrer and Chen Tessler

Want to learn / teach RL? Check out new book draft: Reinforcement Learning - Foundations sites.google.com/view/rlfoundat… W/ shiemannor and Yishay Mansour This is a rigorous first course in RL, based on our teaching at TAU CS and Technion ECE.

For teachers, we also have a 40+ page exam booklet on our website. Why this book? There are several other excellent textbooks, including Sutton and Barto and Bertsekas and Tsitsiklis.

But for teaching RL, we wanted a book that is both rigorous (full proofs, analytical examples), covers what we feel is most relevant, and easy enough for undergrad teaching. The book is a focused one semester course for advanced undergrad/early grad covering key topics in depth.

We hope you find it useful! The book is still work in progress - we’d be grateful for comments, suggestions, omissions, and errors of any kind, at [email protected]

Tired of manual #ComfyUI workflow design? While recent methods predict them, our new paper, FlowRL, introduces a Reinforcement Learning framework that learns to generate complex, novel workflows for you! paper [arxiv.org/abs/2505.21478]